On Sun, Jul 13, 2014 at 12:43 PM, Matthew Petach <mpetach@netflight.com> wrote:
On Sun, Jul 13, 2014 at 10:17 AM, Todd Lyons <tlyons@ivenue.com> wrote:
On Sun, Jul 13, 2014 at 9:53 AM, Matthew Petach <mpetach@netflight.com> wrote: Because that Netflix box is not an on-demand cache, it gets a bunch of shows pushed to it that may or may not be watched by any of Brett's customers. Then the bandwidth he must use to preload that box is large, much larger than the sum of the streams his customers do watch.
However..... (1) There are other considerations besides bandwidth saved: there is customer experience improvement if latency and therefore load times decrease. (2) You or a cache box don't know which streams your customers will watch in advance. Although the cache units preload popular content, not necessarily the entire catalog. Your users are most likely watch during peak hours, which is the time at which more bandwidth is the most expensive... at most other times, additional bandwidth usage is $0, so it doesn't strictly matter, necessarily, if more total transfer is required using a cache box than not. (3) If you don't have at least a couple Gigabits of Netflix traffic, you are unlikely to consider undertaking the expense of the SLA requirements before you can run a box, electricity, space in the first place, if you even meet the traffic minimums required to get free cache boxes. And (4) The "pushing of shows to the units" occur during a configured fill window, which their guides say will be defined by the provider's network planning team in a manner and maximum bandwidth demand over that time suited to your traffic profile, so as to not increase the 95-th percentile traffic from your upstream. For example: the fill window can occur during the hours of the day when there is little interactive customer traffic. They recommend a 10 to 12 hour fill window with a maximum rate of 1.2 Gigabits. http://oc.nflxvideo.net/docs/OpenConnect-Deployment-Guide.pdf Therefore, in any of the cases where cache boxes have actually been implemented properly, they are still likely to be a net benefit for both provider and customers.
Thank you for clarifying that; I thought what Brett was concerned about was traffic in the downstream direction, not traffic for populating the appliance. -- -JH